week 9：Choosing powerful unittest case.

  We sometimes need unittest to help us check whether a code works for all the cases(ideally). For simple codes, we can check it directly by giving a few examples. But for codes that are more complex, like the minimax game strategy we implemented in our assignment, we can not test all the cases, nonetheless, it is reasonable to be confident that our code works perfectly after making it pass some powerful tests.
  Take minimax strategy as an example. Suppose we have current value N for subtract square game state, and for all the 10 possible next moves, 2 out of 10 are the moves that lead to a score 1. So if our suggest_move return one of those two moves, one should expect that it has a pretty low probability that it is just by chance that we get the desired result.
  Furthermore, assume the current value N1 and N2 has n1 and n2 possible independent moves respectively(ignore fact that they may be correlated.) If every time we set case with low probability for computer to randomly choose a right move, theoretically we can make the probability approaching zero by adding enough test cases. But to make it more efficient, we can choose good cases rather than a case that can not differentiate a random suggest_move and minimax. For instance, if there is a state N3, such that every next moves lead to a score -1. Then it can not help to reject the wrong codes.
  In conclusion, we should think wisely to choose powerful test cases that can give strong evidence whether it is a good code.